Introduction

Ovarian cancer is the most lethal disease among gynecological malignancies. The number of patients in Japan with ovarian cancer is increasing constantly. In the past decade, the introduction of various chemotherapeutic agents along with improvements in surgical skills have drastically improved the short-term survival of patients with ovarian cancer. However, long-term survival has not changed significantly because most patients experience repeated recurrence and eventual development of chemo-resistance [1].

Cancer immune therapy has long been expected to become a new modality for solid tumors, including ovarian cancer. Basic research in this field has clarified the principle of cancer immunity during recent decades and, according to these findings, various cancer immune therapies have been suggested and clinical trials on various cancers, including ovarian cancer have been conducted. In reality, however, only very few trials have been clinically successful [2].

One of the reasons for the failure of practical immune therapy is ascribed to the immune-suppressive cancer microenvironment. Recently, it has become clear that cancers often maintain their microenvironment in an immune-suppressive status by using special mechanisms that can be termed generally as “cancer immune escape” strategies. Thus, even if we can induce strong systemic cancer immunity, it may effectively not reach the tumor, leaving it unaffected by the induced anti-cancer immunity [3]. In other words, systemic anti-tumor immunity does not necessarily parallel local immunity; the latter reflects actual cancer immunity and is more important for successful cancer treatment.

Relevance and Significance of Local Immune Environment to the Clinical Course of Ovarian Cancer

PD-ligand (PD-L) molecules were originally identified as being expressed on the surface of antigen presenting cells as a co-factor of immunological synapse; their receptor, PD-1, is expressed mainly on T cells. Such PD-L were known to act as negative regulators of immune reactions by sending suppressive signals to effector T cells. Recently, it has been recognized that some cancers express this protein and, by counter-attacking T cells infiltrating the tumor site, local tumor immunity is weakened [4]. Several years ago, we analyzed the expression of PD-L in ovarian cancer [5]. Using 70 clinical samples of ovarian cancer, we evaluated immunohistochemincal expression of PD-L1 and PD-L2, and correlated them with other clinic-pathological factors and immune factors. High expression of PD-L1 was observed in 68.5% of ovarian cancers, and was not significantly correlated with clinic-pathological factors such as cancer stage, histological subtype and LN metastasis. However, patients with high levels of expression of PD-L1 had significantly poorer prognosis compared with those expressing low levels of PD-L1 (Fig. 1). A multivariate analysis showed that PD-L1 is an independent prognostic indicator.

Fig. 1
figure 1

Prognosis of patients with ovarian cancer according to expression of PD-ligand 1 (PD-L1) or CD8+T cell infiltration. Patient with high expression of PD-L1 have significantly poorer prognosis compared with those with low expression. Also, patients whose cancer shows massive infiltration of CD8+T cells have a significantly better prognosis than patients with less CD8+T cell infiltration

Using the same samples, we also evaluated the number of tumor-infiltrated CD8+T cells, since Zhang et al. reported its relevance to patient outcome [6]. In agreement with their report, ovarian cancer patients whose cancer showed massive infiltration of CD8+T cells had a significantly better prognosis than patients with less CD8+T cell infiltration [5]. Of note, the number of tumor-infiltrated CD8+T cells had a weak but significant inverse correlation with PD-L1 expression, suggesting that PD-L1 expressed in ovarian cancer cells may somehow negatively affect the infiltration of CD8+T cells.

To further investigate the significance of CD8+T cell infiltration in ovarian cancer, we investigated the expression of several other immune-suppressive factors in ovarian cancer and their association with CD8+T cell infiltration. Cyclooxigenase is another immune-suppressive factor that may affect local immune status. The patient group that showed high expression of COX-1 and -2 exhibited low counts of tumor-infiltrated CD8+ T cells [7]. We also investigated the expression of UL-16 binding protein-2 (ULBP2)—another immune-suppressive molecule in ovarian cancer—and found that its expression is a poor prognostic factor. In addition, again, the expression of ULBP2 was negatively correlated with CD8T cell count in the ovarian cancer site [8].

Taken together, these data suggest that the number of CD8+T cells infiltrated into the tumor site is the most reliable predictor of prognosis of ovarian cancer patients, which may well reflect the status of the local immune environment. Multiple factors, including PD-L1, Cox and ULBP2 could prevent the infiltration of CD8+T cells in ovarian cancer and thereby worsen the outcome of the patient.

Activating Local Immunity by CCL19-Transduced Embryonic Endothelial Progenitor Cells in Murine Ovarian Cancer

The results above imply that a therapy that can alter the local immune microenvironment so that more CD8+T cells can infiltrate into the tumor site may become an effective cancer immune therapy. However, advanced cancer usually consists of multiple tumor foci, either in remote metastasis or via peritoneal dissemination. Under such circumstances, if we want to manipulate the immune system to alter the local environment at each focus, we need to develop an efficient system with which to deliver immune-stimulatory signals (Fig. 2). For this purpose, we have investigated the use of endothelial progenitor cells (EPC).

Fig. 2
figure 2

An efficient delivery system is essential to alter local immunity in multiple tumor sites

It is well known that tumors actively construct new vessels when they grow and, in the course of this neovascularization, circulating EPC are actively incorporated into the peripheral blood flow [9]. Recently, an EPC line was established from a mouse embryo (designated eEPC) [10], and the authors kindly provided us with this cell line. EPC have been used to treat ischemic disease in a number of experimental and clinical trials [11]. In the present study, we attempted to utilize the natural property of this cell line to be actively incorporated into tumor vessels as a carrier to deliver the immune-stimulatory signal to the tumor site [12]. Figure 3 shows the schematic concept of our treatment strategy. First, eEPC is transduced with an immune-stimulatory gene to generate immune-stimulatory eEPC. Next, tumor-bearing mice are treated by intravenous injection of transduced immune-stimulatory eEPC. These eEPC become incorporated into the tumor vessels, where they then secrete the immune-stimulatory molecule and alter the local immune environment.

Fig. 3
figure 3

Schematic strategy of our immune-target cancer therapy using embryonic endothelial progenitor cells (eEPC). The eEPC line is transduced with an immune-stimulatory gene to generate immune-stimulatory eEPC. Tumor-bearing mice are then treated by intravenous injection of transduced immune-stimulatory eEPC. The eEPC become incorporated into the tumor vessels, where they secrete immune-stimulatory molecule and alter the local immune environment

A preliminary experiment was first performed in order to select a good candidate as an immune stimulatory molecule from among various immune-modulating signals. The three genes Flt-3 ligand, LIGHT, and CCL19 were selected and retrovirally transduced into eEPC. We used the mouse ovarian cancer cell line HM-1 as a therapeutic model. First, HM-1 cells were mixed with eEPCs transduced with immune-stimulatory genes, which were then inoculated into immune-competent mice to see if they show a tumor inhibitory effect. Among eEPCs, the line transduced with CCL19 was found to be most effective in tumor inhibition. Thus CCL19 was used in further experiments.

CCL19, also known as Exodus-3 or MIP-3-beta, is an immune-stimulatory molecule responsible for the chemotaxis of immune cells and known to facilitate antigen presentation [13, 14]. When we repeated the above experiment with larger numbers of mice, eEPC transduced with CCL19 (designated as eEPC-CCL19) again exhibited significant tumor inhibition. Of note, when the same experiment was performed in immunodeficient (SCID) mice, no apparent tumor inhibitory effect of eEPC-CCL19 was observed, which suggests that the therapeutic effect was mediated by the immune system.

We next assessed the immune status of the tumor local environment in more detail. The infiltration of CD4-, CD8- and CD11c-positive cells in tumor samples was evaluated using immunohistochemincal staining of HM-1 mice inoculated with eEPCs. Only when HM-1 was inoculated with eEPC-CCL19 were numerous immune cells, including CD8+T cells, observed to infiltrate into tumor sites, indicating that the targeted localization of eEPC-CCL19 in the tumor site causes massive immune cell infiltration into the tumor. Moreover, in order to assess systemic immunity, the number of spleen cells producing IFN-g were counted using an ELISPOT assay. Spleen cells harvested from mice in which HM-1 was inoculated with eEPC-CCL19 secreted significantly higher amounts of IFN-g in response to the antigen, indicating that co-injection of eEPC-CCL19 elicited systemic anti-tumor immunity in addition to intensifying the local immunity.

These experiments suggest that if we can co-localize eEPC-CCL19 with HM-1 ovarian cancer cells, they will exert a prominent anti-tumor effect by eliciting strong local, and perhaps also systemic, anti-cancer immunity. So, we next questioned if eEPC-CCL19 is also effective on multiple remote metastases when injected intravenously. Prior to proceeding to an animal therapeutic model, we examined if intravenously injected eEPC can really target tumors. A time course incorporation of EGFP-transduced eEPC was evaluated after immune-fluorescent staining of EGFP and CD31. Injected eEPC was seen to reach the metastatic lung foci, and first extravasate from the vessels. Then, several days later, they were observed to be incorporated into tumor vessels, although numbers were less than expected.

Next, we tested the anti-tumor effect of eEPC-CCL19 in a lung metastasis model (Fig. 4). HM-1 ovarian cancer cells were injected from the tail vain to form lung metastasis. After 4 days, when minute lung metastases were established, eEPC-CCL19s were injected from the tail vain every 2 days for 2 weeks. This treatment markedly diminished the number of lung metastases and significantly prolonged the survival of the tumor-bearing mice.

Fig. 4
figure 4

Effect of eEPC-CCL19 in a lung metastasis model of HM-1 ovarian cancer cells. HM-1 cells were injected from the tail vain to form lung metastasis. After 4 days, eEPC-CCL19s were intravenously injected. This treatment markedly diminished the number of lung metastases and significantly prolonged the survival of tumor-bearing mice

The most common metastatic route of ovarian cancer is peritoneal dissemination. Thus, we next evaluated if eEPC-CCL19 also has a tumor inhibitory effect in the mouse peritoneal dissemination (Fig. 5). HM-1 cells were injected into the mouse peritoneal cavity to induce peritoneal dissemination. On day 4, we started treatment with eEPC-CCL19 injection either via tail vain or by direct intra-peritoneal injection. Interestingly, eEPC-CCL19 was effective only when it was injected intravenously, not directly into the peritoneal cavity. We do not know the exact reason for this difference, but preliminary analysis suggests that eEPCs are incorporated into peritoneal foci much more efficiently when injected intravenously compared with intraperitoneal injection.

Fig. 5
figure 5

Effect of eEPC-CCL19 in a peritoneal dissemination model of HM-1 mouse ovarian cancer cells. HM-1 cells were injected into the mouse peritoneal cavity to induce peritoneal dissemination. On day 4, eEPC-CCL19 was injected either via tail vain or by direct intra-peritoneal injection. eEPC-CCL19 was effective only when injected intravenously

Finally, we tested if eEPC-CCL19 has antitumor effects in another malignant tumor: B16F10 mouse melanoma cells. Again, eEPC-CCL19 showed a significant anti-tumor effect in both subcutaneous injection and lung metastasis models.

Collectively, the results of our study demonstrate that improvement in the local immune environment by intravenous treatment with eEPC-CCL19 may be an effective therapeutic strategy for multiple remote metastases of ovarian cancer.

Conclusions and Future Directions

The local microenvironment of ovarian cancer is affected by various factors, and these may differ from case to case. So in future cancer immune therapy, we should first determine the factors responsible for the “cancer immune escape” in each case, and treatment should be modified accordingly as an individualized immune therapy. For this purpose, the system shown here may be a powerful tool with which to deliver multiple different immune signals sequentially to each cancer nest.

However, to make this approach clinically applicable, many problems need to be solved. One of the most critical problems is the source of the EPC. To avoid immune rejection, we need to use self-bone marrow-derived EPC or at least HLA-matched EPC. One solutions might be the utilization of embryonic stem (ES) cells or induced pluripotent stem cells (iPS cells) [15] as a source of EPC. Recently, a method to differentiate ES cells to EPC has been reported [16] and we are now trying to apply that to our strategy.

Finally, the most important issue when we want to employ immune therapy is to find ways of effectively combining immune therapy with conventional therapies including chemotherapy. Chemotherapy is known to suppress or intensify tumor immunity according to the situation. Therefore, it is especially important to establish a method to evaluate local immune status in each patient in various clinical settings such as post-chemotherapy, post-operation or recurrence.